Layla sat in the glass-walled war room, two decks open. Deck A: “AI-Powered Recommendations — 18% Lift in AOV.” Deck B: “Why the Agentic Checkout Will Backfire.” The CEO had just walked in buzzing from a VC dinner: “Every Series C needs an AI flagship this quarter.”
Every mid-to-senior PM has lived this moment. One slide deck too many, one hallway conversation that could make or break credibility, and the feeling that “no” sounds like career suicide while “yes” without guardrails is professional malpractice.
Convincing For
Stop leading with “AI will transform everything.” Lead with the problem they lose sleep over, then show how AI is the least-bad way to solve it — while being honest about what it won't do.
Helena Tan, AI Agents Lead at Box. Got 30 minutes with a senior leader who opened with: “Document summary stuff is lame. Is this actually going to help me?”
Helena asked: “Walk me through your real Tuesday morning.” Twenty minutes later she had a prototype against actual pain — pulling contract clauses, surfacing risks. The leader took a photo of the screen: “This would actually help me.”
Farah's Triage Win. Support tooling, 10K+ tickets/month. Reps burning out, CSAT tanking. She walked in with: “Stop the Firehose — 25% Deflection Target.”
Showed Trinity explicitly: Data: 65% coverage. Models: 72% accuracy Week 1 — human-in-loop mandatory. UX: Confidence scores, auto-escalate low confidence. Tied to CEO's metric: rep utilization. Deflection hit 27%.
AI Initiative Scorecard
Any Trinity pillar below 7? Science experiment, not a project.
| Business Pain | __ /10 |
| Trinity: Data quality & ownership | __ /10 |
| Trinity: Model feasibility + drift plan | __ /10 |
| Trinity: UX trust mechanisms | __ /10 |
| Hidden Costs (labeling, compute, rollback) | $______ |
| ROI Timeline | “Month 3: ____%” |
| Risks surfaced every 30 days | List: _______ |
| Alternative if killed later | Plan B: ______ |